f_regression#

sklearn.feature_selection.f_regression(X, y, *, center=True, force_finite=True)[source]#

Univariate linear regression tests returning F-statistic and p-values.

Quick linear model for testing the effect of a single regressor, sequentially for many regressors.

This is done in 2 steps:

  1. The cross correlation between each regressor and the target is computed using r_regression as:

    E[(X[:, i] - mean(X[:, i])) * (y - mean(y))] / (std(X[:, i]) * std(y))

  2. It is converted to an F score and then to a p-value.

f_regression is derived from r_regression and will rank features in the same order if all the features are positively correlated with the target.

Note however that contrary to f_regression, r_regression values lie in [-1, 1] and can thus be negative. f_regression is therefore recommended as a feature selection criterion to identify potentially predictive feature for a downstream classifier, irrespective of the sign of the association with the target variable.

Furthermore f_regression returns p-values while r_regression does not.

Read more in the User Guide.

Parameters:
X{array-like, sparse matrix} of shape (n_samples, n_features)

The data matrix.

yarray-like of shape (n_samples,)

The target vector.

centerbool, default=True

Whether or not to center the data matrix X and the target vector y. By default, X and y will be centered.

force_finitebool, default=True

Whether or not to force the F-statistics and associated p-values to be finite. There are two cases where the F-statistic is expected to not be finite:

  • when the target y or some features in X are constant. In this case, the Pearson’s R correlation is not defined leading to obtain np.nan values in the F-statistic and p-value. When force_finite=True, the F-statistic is set to 0.0 and the associated p-value is set to 1.0.

  • when a feature in X is perfectly correlated (or anti-correlated) with the target y. In this case, the F-statistic is expected to be np.inf. When force_finite=True, the F-statistic is set to np.finfo(dtype).max and the associated p-value is set to 0.0.

Added in version 1.1.

Returns:
f_statisticndarray of shape (n_features,)

F-statistic for each feature.

p_valuesndarray of shape (n_features,)

P-values associated with the F-statistic.

See also

r_regression

Pearson’s R between label/feature for regression tasks.

f_classif

ANOVA F-value between label/feature for classification tasks.

chi2

Chi-squared stats of non-negative features for classification tasks.

SelectKBest

Select features based on the k highest scores.

SelectFpr

Select features based on a false positive rate test.

SelectFdr

Select features based on an estimated false discovery rate.

SelectFwe

Select features based on family-wise error rate.

SelectPercentile

Select features based on percentile of the highest scores.

Examples

>>> from sklearn.datasets import make_regression
>>> from sklearn.feature_selection import f_regression
>>> X, y = make_regression(
...     n_samples=50, n_features=3, n_informative=1, noise=1e-4, random_state=42
... )
>>> f_statistic, p_values = f_regression(X, y)
>>> f_statistic
array([1.2...+00, 2.6...+13, 2.6...+00])
>>> p_values
array([2.7..., 1.5..., 1.0...])
On this page
Show Source